def eval():
denoiser.eval()
model.eval()
LR_image = [
join(opt.input, x) for x in listdir(opt.input) if is_image_file(x)
]
SR_image = [
join(opt.Result, x) for x in listdir(opt.input) if is_image_file(x)
]
for i in range(LR_image.__len__()):
t0 = time.time()
LR = Image.open(LR_image[i]).convert('RGB')
with torch.no_grad():
prediction = chop_forward(LR)
t1 = time.time()
print("===> Processing: %s || Timer: %.4f sec." % (str(i), (t1 - t0)))
prediction = prediction * 255.0
prediction = prediction.clamp(0, 255)
Image.fromarray(np.uint8(prediction)).save(SR_image[i])
def eval():
denoiser.eval()
model.eval()
LR_image = [
join(opt.input, x) for x in listdir(opt.input) if is_image_file(x)
]
SR_image = [
join(opt.Result, x) for x in listdir(opt.input) if is_image_file(x)
]
avg_psnr_predicted = 0.0
for i in range(LR_image.__len__()):
t0 = time.time()
LR = Image.open(LR_image[i]).convert('RGB')
LR_90 = LR.transpose(Image.ROTATE_90)
LR_180 = LR.transpose(Image.ROTATE_180)
LR_270 = LR.transpose(Image.ROTATE_270)
LR_f = LR.transpose(Image.FLIP_LEFT_RIGHT)
LR_90f = LR_90.transpose(Image.FLIP_LEFT_RIGHT)
LR_180f = LR_180.transpose(Image.FLIP_LEFT_RIGHT)
LR_270f = LR_270.transpose(Image.FLIP_LEFT_RIGHT)
with torch.no_grad():
pred = chop_forward(LR)
pred_90 = chop_forward(LR_90)
pred_180 = chop_forward(LR_180)
pred_270 = chop_forward(LR_270)
pred_f = chop_forward(LR_f)
pred_90f = chop_forward(LR_90f)
pred_180f = chop_forward(LR_180f)
pred_270f = chop_forward(LR_270f)
pred_90 = np.rot90(pred_90, 3)
pred_180 = np.rot90(pred_180, 2)
pred_270 = np.rot90(pred_270, 1)
pred_f = np.fliplr(pred_f)
pred_90f = np.rot90(np.fliplr(pred_90f), 3)
pred_180f = np.rot90(np.fliplr(pred_180f), 2)
pred_270f = np.rot90(np.fliplr(pred_270f), 1)
prediction = (pred + pred_90 + pred_180 + pred_270 + pred_f +
pred_90f + pred_180f + pred_270f) * 255.0 / 8.0
t1 = time.time()
print("===> Processing: %s || Timer: %.4f sec." % (str(i), (t1 - t0)))
prediction = prediction.clip(0, 255)
Image.fromarray(np.uint8(prediction)).save(SR_image[i])
def load_data(mode='cifar10', batch_size=16):
assert mode in ['cifar10', 'mnist', 'faces'], '未知数据集'
if mode == 'faces':
root_path = 'G:/Dataset/celebAHQ/'
image_list = [x for x in os.listdir(root_path) if is_image_file(x)]
train_list = image_list[:int(0.8 * len(image_list))]
test_list = image_list[int(0.8 * len(image_list)):]
assert len(train_list) > 0
assert len(test_list) >= 0
trainset = MyDataset(train_list,
root_path,
input_height=None,
crop_height=None,
output_height=32,
is_mirror=True)
testset = MyDataset(test_list,
root_path,
input_height=None,
crop_height=None,
output_height=32,
is_mirror=False)
trainloader = MyDataLoader(trainset, batch_size)
testloader = MyDataLoader(testset, batch_size, shuffle=False)
classes = None
return trainset, trainloader, testset, testloader, classes
elif mode == 'cifar10':
root_path = 'G:/Dataset/cifar10/'
classes = ('plane', 'car', 'bird', 'cat', 'deer', 'dog', 'frog',
'horse', 'ship', 'trunk')
transform = transforms.Compose([
transforms.ToTensor(),
transforms.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5))
])
trainset = torchvision.datasets.CIFAR10(root=root_path,
train=True,
download=False,
transform=transform)
testset = torchvision.datasets.CIFAR10(root=root_path,
train=False,
download=False,
transform=transform)
elif mode == 'mnist':
root_path = 'G:/Dataset/mnist/'
classes = ('0', '1', '2', '3', '4', '5', '6', '7', '8', '9')
transform = transforms.Compose(
[transforms.ToTensor(),
transforms.Normalize((0.5, ), (0.5, ))])
trainset = torchvision.datasets.MNIST(root=root_path,
train=True,
download=False,
transform=transform)
testset = torchvision.datasets.MNIST(root=root_path,
train=False,
download=False,
transform=transform)
trainloader = DataLoader(trainset,
batch_size=batch_size,
shuffle=True,
pin_memory=True,
drop_last=False,
num_workers=2)
testloader = DataLoader(trainset,
batch_size=batch_size,
shuffle=True,
pin_memory=True,
drop_last=False,
num_workers=2)
return trainset, trainloader, testset, testloader, classes